1: \documentclass[12pt]{article}

2: \textwidth 16truecm \oddsidemargin-0truecm

3: \usepackage{amssymb}

4: \usepackage{graphicx}

5: \begin{document}

6: \title{Quantum incompressibility and Razumov Stroganov type conjectures}

7: \author {  Vincent Pasquier.}

8: \date{}

9: \maketitle \hskip-6mm { Service de Physique Th\'eorique, C.E.A/

10: Saclay, 91191 Gif-sur-Yvette, France.}

11: 

12: 

13: 

14: 

15: \begin{abstract}

16: We establish a correspondence between polynomial representations of

17: the Temperley and Lieb algebra and certain deformations of the

18: Quantum Hall Effect wave functions. When the deformation parameter

19: is a third root of unity, the representation degenerates and the

20: wave functions coincide with the domain wall boundary condition

21: partition function appearing in the conjecture of A.V. Razumov and

22: Y.G. Stroganov. In particular, this gives a  proof of the

23: identification of the sum of the entries of a $O(n)$ transfer matrix

24: eigenvector and a six vertex-model partition function, alternative

25: to that of P. Di Francesco and P. Zinn-Justin.

26: 

27: \end{abstract}

28: \maketitle

29: 

30: 

31: 

32: 

33: \section{Introduction}

34: 

35: This paper is aimed at establishing a correspondence between the

36: deformation of certain wave functions of the Hall effect and

37: polynomial representations of the Temperley and Lieb (T.L.) algebra.

38: 

39: This work originates from an attempt to understand the conjecture of

40: A.V. Razumov and Y.G. Stroganov

41: \cite{razumov1}\cite{razumov2}\cite{pierce}, and some partial

42: results towards its proof by P.Di Francesco and P. Zinn-Justin

43: \cite{Pdf}.

44: 

45: We consider the analogue of spin singlet wave functions of the Hall

46: effect when one deforms the permutations into the braid group. This

47: amounts to analyze some simple representations of the T.L. algebra

48: on a space of polynomials in $N_e$ variables where $N_e$ is the

49: number of electrons. The relation with the Hall effect arises when

50: we require certain incompressibility properties.

51: 

52: One of the wave functions we consider here is the Halperin wave

53: function \cite{halperin} for a system of spin one half electrons at

54: filling factor two. When the deformation parameter $q$ is a third

55: root of unity, the braid group representation degenerates into a

56: trivial representation. In this way, we obtain a proof alternative

57: to, and apparently simpler than, that given in \cite{Pdf} of the

58: equality between the sum of the components of the transfer matrix

59: eigenvector and the six vertex model partition function with domain

60: wall boundary conditions \cite{razumov2}\cite{korepin}.

61: 

62: Another wave function we consider is the Haldane Rezayi wave

63: function \cite{haldanerezai}\footnote{More precisely a minor

64: modification of it considered in \cite{hallpasq}.} describing a

65: system of electrons of spin one half at filling factor one. This

66: wave function is a permanent, and its deformation is described in

67: terms of Gaudin's determinants \cite{Gaudin}. When $q$ is a third

68: root of unity, it degenerates to the square of the six vertex model

69: partition function.

70: 

71: In a separate publication \cite{pas2}, we shall consider the Moore

72: Read wave function describing spinless bosons at filling factor one

73: \cite{mooreread}. Its deformation involves an extension of the braid

74: group known as the Birman-Wenzl algebra \cite{birman} which can be

75: represented on a polynomial space similarly to the cases presented

76: here. In some appropriate limit, the representation degenerates and

77: the wave function coincides with the transfer matrix eigenvector

78: considered in \cite{Pdf2} related to the conjecture of J.De Gier and

79: B. Nienhuis \cite{degier}.

80: 

81: In general, when a Quantum Hall Effect wave function is discovered,

82: it is soon after observed experimentally. We argue here, that as a

83: bonus, Quantum Hall Effect wave functions and their deformations

84: yield nice mathematical objects. Moreover, all these objects seem to

85: be in relation with striking conjectures emanating from the six

86: vertex model.

87: 

88: \bigskip

89: 

90: Since the permutation group relevant in the quantum Hall effect is

91: technically simpler than the braid group case, let us for

92: pedagogical reasons explain why finding a wave function turns out to

93: be a useful tool to obtain a polynomial representations of the

94: permutation algebra. Essentially, the rest of the paper extends the

95: idea presented here to the braid group case.

96: 

97: We consider electrons in a strong magnetic field projected in the

98: lowest Landau level. In a specific gauge the orbital wave functions

99: are given by:

100: \begin{eqnarray}

101:  \psi_{n}(z)= {z^n\over \sqrt{n!} } e^{- {z \bar z \over 4 l^2}},\label{orbitals}

102: \end{eqnarray}

103: where $z=x+iy$ is the coordinate of the electron, and $l$ the

104: magnetic length defines the length scale related to the strength of

105: the magnetic field. These orbitals are shells of radius $\sqrt

106: {2n}l$ occupying an area $2\pi l^2$. Each orbital $n$ is represented

107: by a monomial $z^n$.

108: 

109: The quantum Hall effect \cite{prange} ground state $\Psi$ is

110: obtained by combining these individual orbitals into a manybody wave

111: function. A monomial $z_1^{\lambda_1}...z_{N_e}^{\lambda_{N_e}}$

112: describes a configuration where the electron $j$ occupies the

113: orbital $\lambda_j$. The wave function is a linear combinations of

114: such monomials. The effect of the interactions is to impose some

115: vanishing properties when electrons are in contact: $\Psi \sim

116: (z_i-z_j)^m$ with $m$ an integer when $z_i-z_j\to 0$.

117: 

118: The physical properties are mainly characterized by the filling

119: factor $\nu$ which is the number of electrons per unit cell of area

120: $2\pi l^2$. When the filling factor is equal to $\nu$, the

121: accessible orbitals and thus the maximal degree in each variable is

122: bounded by  $\nu^{-1} N_e$. On the other hand, the effect of the

123: interactions ($m$) is to force the electrons to occupy more space,

124: thus to occupy higher orbitals and and has the effect of increasing

125: the degree. The problem is thus to obtain wave functions with the

126: maximal possible filling factor (equivalently the lowest degree in

127: each variable) compatible with the vanishing properties imposed by

128: the interactions.

129: 

130: Once such a wave function is obtained, it is the nondegenerate

131: lowest energy state of a Hamiltonian invariant under the

132: permutations, thus we know that it is left invariant under the

133: permutations. By disentangling the coordinate  part from the spin

134: part, we obtain an irreducible representation of the permutation

135: algebra acting on polynomials.

136: 

137: Let us illustrate this point in the case of the Halperin wave

138: function \cite{halperin} which describes a system of spin one half

139: electrons at filling factor two. There are no interactions between

140: the electrons, but due to the Pauli principle, the wave function

141: must vanish when two electrons of the same spin come into contact.

142: Each independent orbital can be occupied with two electrons of

143: opposite spin, which is why the maximal filling factor is equal to

144: two.

145: 

146: An equivalent way to impose the constraint is to require that any

147: linear combination of the spin components of the wave function

148: vanishes when three electrons come into contact. The reason for this

149: is that two of the electrons involved will necessary have the same

150: spin. When this constraint is taken into account with the minimal

151: degree hypotheses, one obtains a space of polynomials which can be

152: recombined with the spin components into a wave function changing

153: sign under the permutations. Thus we know a priori that the spatial

154: part of the wave function carries an irreducible representation of

155: the permutation algebra dual to that of the spins. This is precisely

156: by generalizing this argument to the braid group case that we obtain

157: the representations of the T.L. algebra mentioned above.

158: 

159: In the  permutation group case case, the components have the simple

160: structure of a product of two Slater determinants grouping together

161: the electrons with the same spin and one does not require to

162: recourse to this machinery.

163: 

164: Let us now briefly indicate why the Halperin wave function may have

165: something to do with the eigenvector of a transfer matrix in the

166: link pattern formulation \cite{batchelor}. The wave function is a

167: spin singlet, and the spin components can best be described in a

168: resonating valance bond (RVB) picture as follows: The labels of the

169: electrons are disposed cyclically around a circle and are connected

170: by a link when two electrons form a spin singlet. Links are not

171: allowed to cross in order to avoid overcounting states. These RVB

172: states coincide with the link patterns of \cite{batchelor}. Thus,

173: the Halperin wave function as the eigenvector of the transfer matrix

174: develops on a basis of link patterns. By deforming the permutation

175: action on link patterns into a T.L algebra action, one is forced to

176: deform accordingly the polynomial representation so as to insure the

177: invariance of the total wave function. When $q$ is a third root of

178: unity, this property is shared by the transfer matrix eigenvector

179: and allows to identify the two.

180: 

181: 

182: 

183: 

184: In the braid group case, the situation is technically more involved

185: than for the permutations. Nevertheless, the minimal degree

186: hypothesis combined with some annulation constraint satisfied by

187: linear combination of the spin components yields a wave function

188: with the correct invariance properties. A major difference with the

189: Hall effect is that the cancelation no longer occurs at coincident

190: points, but at points shifted proportionally to the deformation

191: parameter $q$. Typically, we require that for three arbitrary

192: electron labels $i<j<k$ ordered cyclically, the wave function

193: vanishes when the corresponding coordinates take the values $z,\

194: q^2z, q^4z$.

195: 

196:  One is also led to study the

197: affine extension in order to impose cyclic invariance properties

198: which are tautologically satisfied with the permutations. While

199: defined in a natural way on the link patterns, the cyclic properties

200: require to introduce a shift parameter $s$ when we identify the

201: coordinate $i+N_e$ with the coordinate $i$: $z_{i+N_e}=sz_i$. When

202: this shift parameter is related in a specific way to the braid group

203: deformation parameter, the generalized statistics properties can be

204: established coherently. Here, $z_{i+N_e}=q^6 z_i,$ but the same

205: annulation property can also be satisfied with $s$ not related to

206: $q$, and this can be achieved at the price of doubling the degree

207: and enlarging the algebra \cite{pas2}.

208: 

209: 

210: 

211: 

212: 

213: 

214: 

215: In the  the Haldane Rezayi case, \cite{haldanerezai}, the

216: interactions are such that the wave function must vanish as the

217: square of the distance when electrons of the same spin come into

218: contact. For the same reason as before, this amounts to impose that

219: any linear combination of its spin components vanishes as the square

220: of the distance when three electrons come into contact. This wave

221: function is a permanent, and its deformation is described in terms

222: of Gaudin's determinants \cite{Gaudin}. It degenerates to the square

223: of the six vertex model partition function when the deformation

224: parameter is a third root of unity.

225: 

226: 

227: 

228: The paper is organized as follows. In section \ref{hecke-algebra},

229: we recall some properties about Hecke algebras and their polynomial

230: representations. Section \ref{T.L. algebra} introduces the T.L.

231: algebra representation used here. Section \ref{Q.H.E.deformation} is

232: the core of the paper where we work out the deformed Hall effect

233: wave functions.

234: 

235: We have attempted to be self contained, but in order not to overload

236: the text with technicalities, we have relegated most of the proofs

237: to appendices to which we refer when it is useful.

238: 

239: 

240: 

241: 

242: 

243: 

244: \section{Hecke Algebra.\label{hecke-algebra}}

245: 

246: In this section, we recall some known facts about the Hecke and

247: Temperley and Lieb algebras \cite{Pasthese}\cite{Jones}.

248: 

249: The Braid group algebra is generated by the braid group generators

250: $t_{1},t_{2},...,t_{n-1}$, obeying the braid relations:

251: \begin{eqnarray}

252:  t_{i}t_{i+1}t_{i}&=& t_{i+1}t_{i}t_{i+1}\cr

253:  t_{i}t_{j}&=&t_{j}t_i\ \

254: {\rm if} \ |i-j|>1, \label{braid}

255: \end{eqnarray}

256: for $1\le i\le n-1$. It can be convenient to use the notation

257: $t_{ii+1}$ instead of $t_{i}$, and we will use it when necessary.

258: The Hecke algebra is the quotient of the Braid group algebra by the

259: relations:

260: \begin{eqnarray}

261: (t_{i}-q)(t_{i}+{1\over q})=0,\label{hecke}

262: \end{eqnarray}

263: It can also be defined using the projectors $e_{i}=t_{i}-q$ obeying

264: the relations:

265: \begin{eqnarray}

266: e_{i}^2&=&\tau e_{i}, \cr e_ie_j&=&e_je_i\ \ {\rm if} \ |i-j|>1\cr

267: e_{i}e_{i+1}e_{i}-e_{i}&=&

268: e_{i+1}e_{i}e_{i+1}-e_{i+1}.\label{hecke1}

269: \end{eqnarray}

270: where we set $\tau=-(q+q^{-1})$. The Temperley-Lieb (T.L.) algebra

271: $\mathcal{ A}_n$ is the quotient of Hecke algebra by the relations:

272: \begin{eqnarray}

273: e_{i}e_{i+1}e_{i}-e_{i}= e_{i+1}e_{i}e_{i+1}-e_{i+1}=0.\label{T.L.}

274: \end{eqnarray}

275: In $\mathcal{ A}_n$, a trace can be defined \cite{Jones} as:

276: \begin{eqnarray}

277:  {\rm t r}(xe_p)=\tau^{-1} {\rm t r}(x) \ \forall x\in \mathcal{

278:  A}_p.

279:  \label{trace}

280: \end{eqnarray}

281: 

282: The affine Hecke algebra, \cite{BGHP}\cite{Macdo2}\cite{lecture}, is

283: an extension of the Hecke algebra (\ref{hecke}) by generators $y_i,\

284: 1\le i\le n$ obeying the following relations:

285: \begin{eqnarray}

286: &a)&\ \ y_iy_j=y_jy_i\cr &b)&\ \ t_iy_j=y_jt_i \ \ \ {\rm if}\ j\ne

287: i,i+1 \cr &c)&\ \ t_{i}y_{i+1}=y_{i}t_{i}^{-1}\ \ {\rm if} \ i\le

288: n-1.\label{affine-gener}

289: \end{eqnarray}

290: In (\ref{discussion}), we indicate why (\ref{affine-gener}c) is

291: natural from the Yang-Baxter algebra point of view.

292: 

293: This algebra can be endowed with two possible involutions:

294: $e_i^*=e_i,\ y_i^*=y_i^{\pm 1}$, $q^*=q^{\pm 1}$.

295: 

296: The symmetric polynomials in the $y_i$ are central elements.

297: 

298: We define the affine T.L. algebra $\mathcal{ A}_n'$ as the extension

299: of the T.L. algebra (\ref{T.L.}) by the generators $y_i$.

300: 

301: \subsection{Yang's realization of the Affine relations.}

302: 

303: The commutation relations of the affine generators $y_i$ become

304: simpler to understand if we assume that we have a representation of

305: the permutations $k_{ij}$ acting in the natural way on the indices.

306: Let us introduce the operators $x_{ij}=t_{ij}k_{ij}$ for $i<j$ and

307: $x_{ji}=x_{ij}^{-1}$. These operators obey the Yang's relations:

308: \begin{eqnarray}

309: x_{ij}x_{ji}&=&1,\cr  x_{ij}x_{kl}&=&x_{kl}x_{ij}\ \forall i\ne j\ne

310: k\ne l,\cr x_{ij}x_{ik}x_{jk}&=&x_{ik}x_{jk}x_{ij}. \label{yang-op}

311: \end{eqnarray}

312: We also assume that we have commuting operators $s_i$ such that

313: $s_is_jx_{ij}=x_{ij}s_is_j$.

314: 

315: Using (\ref{yang-op}), one verifies that the operators  defined as

316: \cite{BGHP}\cite{lecture}\cite{affine}:

317: \begin{eqnarray}

318: y_1&=&x_{12}x_{13}...x_{1n}s_1\cr

319: y_2&=&x_{23}x_{24}...x_{2n}s_2x_{21}\cr y_n&=&s_n

320: x_{n1}x_{n2}...x_{nn-1} . \label{yang-rep0}

321: \end{eqnarray}

322: commute. Indeed, they coincide with the scattering matrices of Yang

323: \cite{Yang}. (\ref{affine-gener}b) follows directly from

324: (\ref{yang-op}) once we substitute $t_i=x_{ii+1}k_{ii+1}$.

325: (\ref{affine-gener}c) is a direct consequence of the definition

326: (\ref{yang-rep0}) of $y_i$.

327: 

328: Gathering the permutation operators $k_{ij}$ together, we can obtain

329: another presentation of the $y_i$. Let us introduce the cyclic

330: operator:

331: \begin{eqnarray}

332: \sigma=k_{n-1n}...k_{23}k_{12}s_1.\label{sigma-k}

333: \end{eqnarray}

334: Then we have:

335: \begin{eqnarray}

336: y_1&=&t_{1}t_{2}...t_{n-1}\sigma \cr

337: y_2&=&t_{1}^{-1}y_1t_{1}^{-1}\cr

338: y_n&=&t_{n-1}^{-1}y_{n-1}t_{n-1}^{-1}.\label{affine-rep1}

339: \end{eqnarray}

340: 

341: We can define an additional generator to the $t_{i}$: $t_{n}=\sigma

342: t_{1}\sigma ^{-1}$, which makes the relations (\ref{braid}) become

343: cyclic. One has:

344: \begin{eqnarray}

345: \sigma t_i=t_{i-1}\sigma .\label{affine-rep1-cycle}

346: \end{eqnarray}

347: So that the affine Hecke (or T.L.) algebra is generated by the

348: generators $t_i$ and the cyclic operator $\sigma$ obeying

349: (\ref{affine-rep1-cycle}) and does not require a representation of

350: the permutations. $\sigma^n$ is a central element which can be set

351: equal to one, and $\sigma^*=\sigma^{-1}$ if we take $t^*=t^{-1}$.

352: 

353: Given the Hecke algebra, there is a simple realization of the affine

354: Hecke algebra which consists in taking $y_1=1$. Then, $\sigma$ is

355: defined as:

356: \begin{eqnarray}

357: \sigma=  t_{n-1}^{-1}...t_{1}^{-1} .\label{cycle}

358: \end{eqnarray}

359: 

360: Using the braid relations, one sees that $\sigma t_i=t_{i-1}\sigma$

361: for $ i>1$, and one can define  $t_n$  by $t_n=\sigma

362: t_{1}\sigma^{-1} $. Using the braid relations again, one gets

363: $\sigma t_n=t_{n-1}\sigma $. This defines an operator $\sigma $

364: which allows to construct the affine generators with

365: (\ref{affine-rep1}).

366: 

367: 

368: 

369: \subsection{Polynomial representations.\label{Polynomial

370: representations of the Hecke algebra}}

371: 

372: We consider a space of polynomials in $n$ variables $z_i,$

373: constructed as a linear combinations of monomials:

374: $z^{\mu}=z_1^{\mu_1}z_2^{\mu_2}...z_n^{\mu_n} $ with a total degree

375: $|\mu|=\sum \mu_i$ fixed. There is a natural action of the

376: permutations and of the operators $s_i$ on this space defined by:

377: \begin{eqnarray}

378: \bar\psi(z_1,..z_i..z_j..,z_n)k_{ij}&=&

379: \bar\psi(z_1,..z_j..z_i..,z_n),\cr

380: \bar\psi(z_1,..,z_i..,z_n)s_{i}&=& c

381: \bar\psi(z_1..,sz_i..,z_n).\label{polyaction}

382: \end{eqnarray}

383: It is convenient to consider the polynomials in an infinite set of

384: variables $z_i,\ i\in \cal Z,$ with the identification: $z_{i+n}=s

385: z_i$. The operator $\bar \sigma$ (\ref{sigma-k}) takes the form:

386: \begin{eqnarray}

387: \bar\psi\bar \sigma(z_i)=c\bar\psi(z_{i+1}).\label{sigmacycle1}

388: \end{eqnarray}

389:  The condition $\sigma^n=1$ imposes the relation $c^n

390: s^{|\mu|}=1$.

391: 

392:  As shown in (\ref{poly-hecke}), it is straightforward to derive the

393: following representation of the Hecke relations

394: (\ref{braid},\ref{hecke}):

395: \begin{eqnarray}

396: \bar t_{ij}=-q^{-1}+(1-k_{ij}){qz_i-q^{-1}z_j\over

397: z_i-z_j}.\label{polyactionTL}

398: \end{eqnarray}

399: In this way we obtain a representation of the affine Hecke algebra

400: acting on homogenous polynomials of a given total degree.

401: 

402: The operators $x_{ij}$ take the form:

403: \begin{eqnarray}

404: x_{ij}=-q^{-1}+(q-q^{-1})(1-k_{ij}){z_j\over

405: z_i-z_j}.\label{polyactionX}

406: \end{eqnarray}

407: 

408: In the appendix \ref{diag-yi}, we show that there is a natural order

409: on the monomial basis, $z^{\mu}$, for which the operators $x_{ij}$,

410: and hence the $y_i$ are realized as lower triangular matrices.

411: 

412: The operator $y=y_1+...+y_n$ can be seen to commute with the Hecke

413: generators. It is therefore equal to a constant in an irreducible

414: representation. Its eigenvalue evaluated on the highest weight

415: polynomial $P_{\lambda}$ thus characterizes the representation. It

416: is given by:

417: \begin{eqnarray}

418: y_{\lambda}=c(-q)^{1-n}

419: (s^{\lambda_1}+s^{\lambda_2}q^2+...+s^{\lambda_n}q^{2(n-1)}).

420: \label{valpropY}

421: \end{eqnarray}

422: 

423:  If $\lambda'$ is a permutation of the partition $\lambda$, we say that

424: $z^{\lambda'}$ is of degree $\lambda$. In this paper, we are mainly

425: concerned with the monomials $z^{\lambda'}$, of degree:

426: \begin{eqnarray}

427: \lambda=({n\over 2}-1,{n\over 2}-1,{n\over 2}-2,{n\over

428: 2}-2,...,0,0), \label{partition}

429: \end{eqnarray}

430: and of total degree $|\lambda|={n\over 2}({n\over 2}-1)$.

431: 

432: 

433: We will consider the subclass $\lambda_{\pi}$ of permutations of

434: $\lambda$ which are smaller than $\lambda$ for the order introduced

435: in the appendix \ref{diag-yi}. According to the analysis made there,

436: $\lambda_\pi$ can be obtained from $\lambda$ by a sequence of

437: permutations $(\lambda'_i,\lambda'_{i+1})\to

438: (\lambda'_{i+1},\lambda'_{i})$ with $\lambda'_i>\lambda'_{i+1}$. The

439: only monomials which can be obtained that way are indexed by the

440: standard Young tableaus with two columns of ${n\over 2}$ boxes:

441: \begin{eqnarray}

442:  z^{\lambda_{\pi}}=(z_{\mu_1}z_{\nu_1})^{n\over

443: 2}(z_{\mu_2}z_{\nu_2})^{{n\over 2}-1}...(z_{\mu_{n\over

444: 2}}z_{\nu_{n\over 2}})^0,\label{youngtableau}

445: \end{eqnarray}

446: with $\mu_1>\mu_2>...>\mu_{n\over 2} $, $\nu_1>\nu_2>...>\nu_{n\over

447: 2} $, and $\mu_i>\nu_i$. To simplify notations, we denote these

448: monomials by $z^{\pi}$ instead of $z^{\lambda_\pi}$.

449: 

450: We identify the standard Young tableaus with the paths $\pi=[h_i]$

451: introduced in the appendix \ref{wordrepresentation}: $h_0=h_{n}=0$,

452: $h_i\ge 0$ and $h_{i+1}-h_{i}=\pm 1$. These paths are obtained using

453: the rule: $h_i-h_{i-1}=1$ if $i\in \{\mu_j\}$, and $h_i-h_{i-1}=-1$

454: if $i\in \{\nu_j\}$. For the paths, we use the order $\pi\ge\pi'$,

455: if $[h_i]\ge[h'_i]\ \forall i,$ which coincides with the reverse

456: order for the monomials: $z^{\pi}\le z^{\pi'}$.

457: 

458: This identification is illustrated in figure

459: \ref{fig:wordrepresentation}.

460: 

461: 

462: 

463: 

464: 

465: \section{Representation of the affine T.L. algebra on words.\label{T.L. algebra}}

466: 

467: For $n$ even, there is a simple representation $(\mathcal{H}_n)$ of

468: the T.L. algebra $\mathcal{ A}_n$ obtained as follows. One considers

469: the left action of $\mathcal{ A}_n$ on the space $\mathcal{

470: A}_n\alpha$ where $\alpha$ is the minimal projector $\alpha

471: =e_{1}e_{3}...e_{n-1}$. A basis of this space is given by reduced

472: monomial words in the $e_i$. The elements of this basis can be put

473: into correspondence with paths or link patterns. In the appendix

474: \ref{wordrepresentation} we exhibit a basis of reduced words and we

475: define an order relation on the reduced words.

476: 

477:  A scalar product can be defined as:

478: \begin{eqnarray} \pi^* {\pi'}=\langle \pi|\pi'\rangle \alpha,

479: \label{produitscalaire}

480: \end{eqnarray}

481: where $e^*_i=e_i$ and the involution reverses the order of the

482: letters. In the link-pattern representation, this scalar product is

483: given by: $\tau^l$ where $l$ is the number of loops one gets by

484: stacking the link patterns of $\pi$ and $\pi'$ on top of each other.

485: If $\tau = -(q+{q^{-1}})$ with $q$ not a root of one, this scalar

486: product is positively definite \cite{Jones}. For this scalar product

487: the T.L generators $e_i$ are by construction hermitian.

488: 

489: To obtain the affine algebra representation, let us define as in

490: (\ref{cycle}) the cyclic operator:

491: \begin{eqnarray}

492: \sigma=-q^{{n\over 2}-2} t^{-1}_{n-1}...t^{-1}_{1}, \label{cycle1}

493: \end{eqnarray}

494: where the normalization is such that in the link-pattern

495: representation, $\sigma $ acts by cyclicly permuting the indices

496: $i\to i-1$ (see appendix \ref{actionofsigmaonwords}). One can define

497: an additional generator, $e_{n1}=\sigma e_{12} \sigma^{-1}$, which

498: acts in the same way as $e_{ii+1}$ with the two indices $1,n$. The

499: affine generators are constructed using (\ref{affine-rep1}) with

500: $y_1=-q^{{n\over 2}-2}$.

501: 

502: In the appendix \ref{triangularity}, we show that the operators

503: $y_i$ are realized as triangular matrices in $\mathcal{H}_n$, they

504: are hermitian for the choice $q=q^*$. Their sum $y=\sum_i y_i$ is

505: constant with a value given by:

506: \begin{eqnarray}

507: y=-(q+{q^{-1}}){q^{n\over 2}-q^{-{n\over 2}}\over q-q^{-1}}.

508: \label{valpropy}

509: \end{eqnarray}

510: 

511: There is an imbedding of $\mathcal{H}_{n-2}$ into $\mathcal{H}_{n}$

512: given by $\pi \to \pi e_{1}$ and a projection $E$ from

513: $\mathcal{H}_{n}$ to $\mathcal{H}_{n-2}$ given by:

514: \begin{eqnarray}

515: e_1\pi=\tau E(\pi)e_1 .\label{project-2}

516: \end{eqnarray}

517: This projection is both triangular and hermitian.

518: 

519: In \ref{identify}, we identify $\mathcal{H}_n$ with $\mathcal{

520: A}_{n\over 2}$. This allows us to interpret the projection $E$ as a

521: conditional expectation value of $\mathcal{ A}_{n\over

522: 2}\to\mathcal{ A}_{{n\over 2}-1}$ \cite{Jones}.

523: 

524: \bigskip

525: \section{q-deformed Quantum Hall Effect wave functions.\label{Q.H.E.deformation}}

526: 

527: \subsection{Statement of the Problem.}

528: 

529: Let us consider a vector $\Psi$:

530: \begin{eqnarray}

531: \Psi=\sum_{\pi}\pi F_{\pi}(z_i), \label{PDF}

532: \end{eqnarray}

533: constructed in the following way. The vectors $\pi$ are the basis

534: vectors of $\mathcal{H}_n$ on which the  T.L. algebra acts to the

535: left. $F_{\pi}$ are homogeneous polynomials in the variables

536: $z_1,z_2,...,z_n$ ($n$ is even). The polynomial coefficients of

537: $\Psi$ carry a representation of the affine Hecke algebra generated

538: by the operators $\bar t_i$ and $\bar \sigma $ acting to the right.

539: The problem is to determine the coefficients $F_\pi$ in such a way

540: that both actions give the same result on the vector $\Psi$:

541: \begin{eqnarray}

542: \Psi\bar t_{i}&=& t_{i}\Psi \cr\Psi \bar\sigma&=&

543: \sigma\Psi,\label{duality}

544: \end{eqnarray}

545: The first of these relations is equivalent to the more familiar

546: relation (\ref{Y=k}) derived in \ref{poly-hecke}.

547: 

548: Said differently, we look for a dual action of the affine T.L.

549: algebra acting on polynomials. Unless we specify it, we address this

550: problem for a generic value of the parameter $q$, not a root of

551: unity, for which the T.L. algebra is semisimple \cite{Jones}.

552: 

553: 

554: \subsection{ Module $\mathcal{M}_n$.}

555: 

556: The dual representation of $\mathcal{H}_n$ is obtained by acting

557: with the T.L. generators on the dual $F_\omega$ of the highest

558: vector $\omega \in \mathcal{H}_n$. $\omega$ is given by the sequence

559: $(a_{2p+1}=p+1)$ in the characterization of words we use in the

560: appendix \ref{wordrepresentation} and is fully characterized by the

561: property that it can be written $\omega=e_i \pi$ only for $i={n\over

562: 2}$. The dual vector $F_{\omega}$ must therefore be annihilated by

563: all the $e_i$ with $i\ne n/2$. We realize the module $\mathcal{H}_n$

564: upon acting on $F_{\omega}$ with the generators $e_i$ for $1\le i\le

565: n-1$. We define:

566: \begin{eqnarray}

567: \mathcal{M}_n=\mathbf{ Vec}

568: \{\bar\psi=F_{\omega}\psi\},\label{moduleH'}

569: \end{eqnarray}

570: where we denote with a bar $\bar \psi$ the result of the action of

571: the monomial $\psi$ on $F_{\omega}$. Thus we have $\bar

572: 1=F_{\omega}$. In the appendix \ref{formedehecke}, we show that

573: $\mathcal{M}_n$ defined in this way is a module over the T.L.

574: algebra as long as the $e_i$ obey the Hecke relations

575: (\ref{hecke1}). In other words, the projectors

576: $U^-_{i,i+1}=e_ie_{i+1}e_i-e_i$ are null in $\mathcal{M}_n$. This

577: formal module is however not isomorphic to $\mathcal{H}_n$ unless

578: $F_{\omega}$ obeys some supplementary condition (\ref{fock}). Here,

579: we construct a representation of the T.L. algebra by identifying a

580: state $F_{\omega}$ dual to $\omega$ and satisfying the condition

581: (\ref{fock}).

582: 

583: \smallskip

584: 

585: The expression of the T.L. generators $e_i=t_i-q$ for $1\le i\le

586: n-1$  follows from (\ref{polyactionTL}):

587: \begin{eqnarray}

588: e_i&=&-{qz_{i+1}-q^{-1}z_{i}\over z_{i+1}-z_{i}}(1+k_{ii+1})\cr

589: e_i-\tau&=&(1-k_{ii+1}){qz_{i}-q^{-1}z_{i+1}\over z_i-z_{i+1}}.

590: \label{polynomeTLei}

591: \end{eqnarray}

592: The effect of  $e_i$ and $\tau-e_i$ is to split a polynomial

593: $\bar\psi$ into two polynomials belonging to $\mathcal{M}_n$, $\bar

594: \psi=S_1+(qz_i-q^{-1}z_{i+1})S_2$, where both $S_1$ and $S_2$ are

595: symmetrical under the exchange of $z_i$ and $z_{i+1}$. This

596: decomposition is unique and characterizes the projector $e_i$.

597: 

598: It can be convenient to distinguish the representation on

599: $\mathcal{H}_n$ from its dual on $\mathcal{M}_n$. When this is the

600: case, we denote $\bar e_i$ the dual projectors which act on

601: polynomials.

602: \smallskip

603: 

604:  One  verifies that:

605: \begin{eqnarray}

606: \Delta_{m}^{q}(z_1,...,z_p)=\prod_{1\le i<j \le m} (q z_i-q^{-1}

607: z_j) \label{deltadef}

608: \end{eqnarray}

609: is annihilated by all the $e_i,\ 1\le i\le m-1$, and this defines

610: $\Delta_{m}$ up to a  product by a symmetric polynomial in

611: $z_1,...,z_m.$ Therefore, the minimal degree polynomial candidate

612: for $F_{\omega}$ is:

613: \begin{eqnarray}

614: F_{\omega}=\Delta_{n\over 2}^q(z_1,...,z_{n\over 2}) \Delta_{n\over

615: 2}^q(z_{{n\over 2}+1},...,z_n).\label{Fomegabar}

616: \end{eqnarray}

617: This polynomial cannot be q-antisymmetrized over  ${n\over 2}+1$

618: variables, since the result would have a degree at least ${n\over

619: 2}$ in $z_1$, and this is the content of the condition (\ref{fock}).

620: Thus, $\mathcal{M}_n$ is a simple module which can be identified

621: with $\mathcal{H}_n$. This representation is characterized by its

622: Young diagram $(2^{n\over 2})$ having two columns of length ${n\over

623: 2}$. Its dimension is given by the Catalan number $C_n= {n\choose

624: {n\over 2}}-{n\choose {n\over 2}-1}$.

625: 

626: \smallskip

627: 

628: 

629: If we denote $\pi|_{\omega}$ the coefficient of $\omega$ in the

630: reduced expression of $\pi$, we can identify the polynomials

631: $\bar\psi\in \mathcal{M}_n$ with the dual of $\mathcal{H}_n$ through

632: the relation: $\bar\psi(\pi)=\psi\pi|_\omega$.

633: 

634: We can also introduce the dual basis $F_\pi$ defined by its action

635: on reduced words:

636: \begin{eqnarray}

637: F_\pi(\pi')=\delta_{\pi,\pi'}.\label{vraiedualite}

638: \end{eqnarray}

639: 

640: Let $\pi_{\psi}$ be the complementary word of $\psi$ (defined in

641: appendix \ref{module}) such that one can write

642: $\psi\pi_{\psi}=\omega$ without reducing the expression. One has

643: $\psi\pi_{\psi}|_{\omega}=1$ and $\psi \pi|_{\omega}=0$ if

644: $\pi<\pi_{\psi}$ . Expanding $\bar\psi$ on the basis $F_{\pi}$, we

645: get: $\bar\psi=\sum_{\pi\ge \pi_\psi} \psi\pi|_{\omega}F_{\pi}$, and

646: by inverting the triangular system, we can obtain the expression of

647: $F_\pi$.

648: 

649: \smallskip

650: 

651:  Let us verify that the highest monomial of $\bar\psi$, and thus of

652: $F_{\pi_\psi}$ as well, is proportional to $z^{\pi_{\psi}}$. We show

653: this by recursion. It is true for $\psi=1$: $\bar 1=F_{\omega}$ ,

654: $\pi_1=\omega$ and $z^{\omega}$ is the highest monomial of

655: $F_{\omega}$. We assume that the property is true for $\psi'<\psi$.

656: If $\psi\ne1$, one can write $\bar\psi=\bar\psi' e_i$ with

657: $\bar\psi'<\bar\psi$, and we have $\pi_{\psi'}=e_i \pi_{\psi}$ with

658: $\pi_{\psi}<\pi_{\psi'}$.

659: 

660: Then,  according to the recursion hypothesis, the highest monomial

661: of $\bar \psi'$ is $z^{\pi_{\psi'}}$ which  contains the factor

662: $z_i^{m}z_{i+1}^n$ with $n>m$. Since $z_i^{m}z_{i+1}^n \bar

663: e_i=-qz_i^{n}z_{i+1}^m+ {\rm lower\ monomials},$ the highest

664: monomial of $\bar\psi=\bar\psi'\bar e_i$ is $z^{\pi_{\psi}}.$

665: 

666: We also obtain the normalization coefficient of $z^\pi$ up to a

667: global factor: $F_\pi=c_\pi z^\pi+{\rm lower \ monomials},$ with

668: $c_\pi= (-{1\over q})^{l_{\pi}}$, and $l_{\pi}$ is the number of

669: letters $e_i$ entering the reduced expression of $ \pi$.

670: 

671: 

672: 

673: \bigskip

674: \subsection{ Module $\mathcal{M}_n'$.\label{moduleM'}}

675: 

676: 

677: 

678: We now consider a larger module $\mathcal{M}_n' \supset

679: \mathcal{M}_n$ by letting the operator $\bar\sigma$ defined in

680: (\ref{sigmacycle1}) act on the polynomials. We will put some

681: constraint on the parameter $s$ (which characterizes $\bar\sigma$)

682: to have $\mathcal{M}_n'=\mathcal{M}_n$. We consider the simple case

683: $n=4$ in the appendix \ref{H4} and we obtain $s=q^6$ which is the

684: general case as we show here.

685: 

686: 

687: Let us assume that $\mathcal{M}_n'=\mathcal{M}_n$ and see what

688: constraints $s$ must satisfy to identify $\bar\sigma$ defined by its

689: action on polynomials (\ref{sigmacycle1}) with $\sigma$ defined in

690: terms of generators (\ref{cycle}).

691: 

692:  We observe that $\sigma^{-1}\omega_n=e_1\omega_{n-2}$, where

693: $\omega_{n-2}$ is the highest state in $\mathcal{H}_{n-2}$. This can

694: easily be verified in the link pattern representation. Thus, we must

695: have:

696: \begin{eqnarray}

697: \sigma E(\pi)e_1|_{\omega_n}=

698: E(\pi)e_1|_{\sigma^{-1}\omega_n}=E(\pi)|_{\omega_{n-2}}.\label{chaineduale}

699: \end{eqnarray}

700: 

701:  Let us consider the dual to the projection $E$, $E'$ from $\mathcal{M}_n\to \mathcal{M}_{n-2}$

702: defined as $\bar \psi e_1=\tau E'(\bar \psi).$ $E'$ needs to satisfy

703: the conditions:

704: \begin{eqnarray}

705:  &a)& \ \ E'( \bar\psi e_1)=\tau E'(\bar \psi)\cr

706:  &b)& \ \ E'( \bar\psi e_i)=E'( \bar\psi)e_i\ \forall i>2\cr

707:  &c)&\ \ E'(\bar\psi e_1)=0\Rightarrow \bar\psi e_1=0.

708:  \label{condition1E'}

709: \end{eqnarray}

710: >From (\ref{chaineduale}), in order to identify $\bar \sigma$ with

711: $\sigma$, we see that the projection $E'$ must satisfy:

712: \begin{eqnarray}

713: E'(F_{\omega_n}\bar\sigma)=F_{\omega_{n-2}}.\label{condition2E'}

714: \end{eqnarray}

715: $E'$ can be realized  as:

716: \begin{eqnarray}

717: E'(\bar \psi)=c'{1\over \phi(z,z_i)}\bar

718: \psi(z_1={z},z_2=q^2z,z_i),\label{definitE'}

719: \end{eqnarray}

720: where $\phi(z,z_i)=\prod_{i=3}^{n}(z_i-q^4z)$ and $c'$ is a

721: normalization constant. $E'$ verifies (\ref{condition1E'}a,b) by

722: construction as can be seen from the expression (\ref{polynomeTLei})

723: of $e_1-\tau$.

724: 

725: Using the explicit expression (\ref{sigmacycle1}) of $\sigma$, we

726: have:

727: \begin{eqnarray}

728: E'(F_{\omega_n}\sigma)=c' s^{-{n\over 4}-{1\over 2}}{1\over

729: \phi(z,z_i)}\prod_3^{{n\over 2}+1}(q^3z-q^{-1}z_i)\prod_{{n\over

730: 2}+2}^n(q z_i-q^{-1}s

731: z)F_{\omega_{n-2}}(z_3,...,z_n).\label{definitE1'}

732: \end{eqnarray}

733: which imposes $s=q^6$ for the polynomial in the numerator to be

734: proportional to $\phi(z,z_i)$ and (\ref{condition2E'}) to be

735: satisfied.

736: 

737: \bigskip

738: 

739: To identify $\mathcal{M}_n$ and $\mathcal{M}_n'$, we give a more

740: convenient characterization of $\mathcal{M}_n'$. Consider the space

741: $\mathcal{M}_n''$ of homogenous polynomials in $n$ variables, and of

742: the minimal total degree, obeying the property:

743: \begin{eqnarray}

744:  {\rm(P):}\ \ \bar\psi(z_i=z,z_j=q^2z,z_k=q^4z)=0,\ \ {\rm if}\ i,j,k,\ {\rm are\ cyclically\ ordered}.

745:  \label{propiete}

746: \end{eqnarray}

747:  This property is obviously compatible with the cyclic

748: identification $z_{i+n}=q^6z_i$, it is thus preserved by

749: $\bar\sigma$ (\ref{sigmacycle1}). By applying (P) to the triplets

750: $(1,2,j)$, we see that the projection (\ref{definitE'}) is well

751: defined from $\mathcal{M}_n''$ to $\mathcal{M}_{n-2}''$.

752: 

753: 

754: \bigskip

755: 

756: We show that $\mathcal{M}_n''=\mathcal{M}_n$. For this, we first

757: show that $\mathcal{M}_n''$ is a module over the T.L. algebra

758: $\mathcal{A}_n$ and that it contains $\mathcal{M}_n$, then we show

759: that $\mathcal{M}_n''$ is irreducible over $\mathcal{A}_n$.

760: 

761: \smallskip

762:  To show that $\mathcal{M}_n''$ is a module over $\mathcal{A}_n$, we

763: verify that the generators $e_i$ preserve the property (P). Assuming

764: that the polynomial $\bar \psi$ verifies (P) we verify that

765: $\bar\psi e_i$ obeys (P) for a cyclically ordered triplet $k,l,m$.

766: If $\{i,i+1\}\cap \{k,l,m\}=\emptyset $, it is obvious. If $i+1=k$,

767: it results from the fact that $\bar\psi$ obeys (P) for the triplets

768: $i,l,m$ and $i+1,l,m$. The same type of argument applies if $i=m$.

769: If $\{i,i+1\}\subset \{k,l,m\} $, $\bar\psi(e_i-\tau)$ is

770: proportional to $(qz_i-q^{-1}z_{i+1})$ and therefore obeys (P).

771: 

772: 

773: 

774: \smallskip

775: 

776: Let us show that (\ref{condition1E'}c) is satisfied in

777: $\mathcal{M}_n''$. If $E'(\bar\psi e_1)=0$, $\bar\psi e_1$ vanishes

778: when $z_2=q^2z_1$, and from the definition (\ref{polynomeTLei}) of

779: $e_1$, it  is symmetric in $z_1,z_2$. It is therefore divisible by

780: $(z_1-q^2z_2)(z_2-q^2z_1)$. Hence, $\bar\psi e_1/(z_1-q^2z_2)$

781: satisfies (P) and has a total degree reduced by one. It is thus

782: equal to zero according to our minimal degree hypothesis.

783: 

784: \smallskip

785: 

786: It is clear that $F_{\omega}$ satisfies the property (P). To show

787: that $\mathcal{M}_n\subset \mathcal{M}_n''$, we need to show that

788: the degree of the polynomials in $\mathcal{M}_n''$ is the degree

789: ${n\over 2}({n\over 2}-1)$ of $F_\omega$. We proceed by recursion on

790: $n$ and for the moment, we exclude the case where $e_1$ is

791: represented as zero in $\mathcal{M}_n''$. Due to

792: (\ref{condition1E'}c) there are polynomials $\bar\psi$ in

793: $\mathcal{M}_n''$ such that $E'(\bar\psi)\ne0$. This implies that

794: $\bar\psi$ has a degree at least $n-2$ in $z_1,z_2$. We can apply

795: the recursion hypothesis to $E'(\bar\psi)\in \mathcal{M}_{n-2}''$ to

796: conclude that the minimal degree is ${n\over 2}({n\over

797: 2}-1)$.\footnote{The same argument shows that the maximal degree of

798: the polynomials in $\mathcal{M}_n''$ is  $\ge \lambda$ for the order

799: defined in \ref{diag-yi}.}

800: 

801: 

802: 

803: \smallskip

804: 

805: To show that $\mathcal{M}_n''$ is irreducible as a T.L. module, we

806: use the recursion hypothesis that

807: $\mathcal{M}_{n-2}''=\mathcal{M}_{n-2}$. Due to

808: (\ref{condition1E'}c), $E'$ is injective from $\mathcal{M}_n''e_1$

809: to $E'(\mathcal{M}_n'')\subset \mathcal{M}_{n-2}$. Since

810: $\mathcal{M}_ne_1=\mathcal{M}_{n-2}\subset \mathcal{M}_n''e_1$, we

811: have $\mathcal{M}_n''e_1=\mathcal{M}_ne_1$. Thus, if

812: $\mathcal{M}_n''$ contains an irreducible submodule $R\ne

813: \mathcal{M}_n$, $Re_1=0$. If $Re_1=0,$ from (\ref{T.L.}) we see that

814: all the $e_i$ are represented as $0$ in $R$, and therefore, the

815: polynomials in $R$ are proportional to $\Delta_{n}$ defined in

816: (\ref{deltadef}) times a symmetric polynomial. Since the total

817: degree of $\Delta_n$ is larger than ${n\over 2}({n\over 2}-1)$,

818: $R=0$. We conclude that $\mathcal{M}_n''=\mathcal{M}_n$ as a T.L.

819: module.

820: 

821: \smallskip

822: 

823: Finally, to identify $\mathcal{M}_n''$ and $\mathcal{M}_n$ as affine

824: modules, we observe that $y_1=\sigma^{-1}\bar\sigma$ commutes with

825: $\mathcal{A}_{n-1}$ generated by $e_2,...,e_n$. Since

826: $\mathcal{M}_n$ is irreducible over $\mathcal{A}_{n-1}$

827: \cite{Jones}, $y_1$ is proportional to the identity, thus $\sigma$

828: and $\bar\sigma$ can be identified.

829: 

830: 

831: \bigskip

832: \subsubsection{ Relation with the Macdonald Polynomials and the work of Di Francesco and Zinn-Justin.}

833: 

834:  As a check of consistency, we must verify that the two expressions

835: of the eigenvalue of the central operators $y$

836: (\ref{valpropY},\ref{valpropy}) are the same when $s=q^6$. This is

837: indeed the case if we substitute in  (\ref{valpropY}) the degree

838: $\lambda$ (\ref{partition}) of the highest polynomial in

839: $\mathcal{M}_n$ and $c=q^{3(1-{n\over 2})}$.

840: 

841: For a generic $s$, the operator $y$ (\ref{valpropY}) can be

842: diagonalized on the basis of symmetric polynomials and its

843: eigenvectors define the Macdonald polynomials \cite{Macdo1}. We have

844: seen that when $s=q^6$, the polynomial representation is reducible.

845: As a counterpart, some diagonal elements  $y_{\lambda'}$ of $y$

846: become degenerate with $y_\lambda$, for example,

847: $\lambda'_2=\lambda_2-1$, $\lambda'_5=\lambda_5+1$. Thus, $y$ cannot

848: be diagonalized. We must use another operator such as ${dy\over ds}$

849: to define the analogous symmetric polynomial.

850: 

851: In the non semisimple case $q^2+q+1=0,\ (\tau=1),$ the T.L.

852: representation admits a sub-representation given by ${\rm Vec}

853: \{\sum x_\pi \pi, {\rm\ with} \ \sum x_\pi=0\}.$ The trivial

854: representation $\Omega$ is obtained by equating to zero these

855: vectors. The dual polynomial $F_{\Omega}=\sum_{\pi}F_\pi$ is

856: therefore symmetrical of degree $\lambda$, and obeys the property

857: (P)(\ref{propiete}). This completely determines it to be

858: proportional to the Schur function $s_{\lambda}$ with $\lambda$

859: given by (\ref{partition}). Indeed, $s_{\lambda}$ has a degree

860: $\lambda$ and satisfies (P) since three columns of the determinant

861: which defines it become linearly dependant when we make the

862: substitution (P). By the same argument as used in \ref{moduleM'},

863: the degree of a symmetric polynomial satisfying (P) must be at least

864: $\lambda$ (relatively to the order of partitions which follows from

865: \ref{diag-yi}), and this proves its uniqueness.

866: 

867: In this limit, the $F_\pi$ are also the components the ground state

868: of the $O(n=1)$ transfer matrix, and thus, we have proved that this

869: sum is $s_{\lambda}$, which is the result of \cite{Pdf}.

870: 

871: It would be interesting to see if in this limit, $F_{\Omega}$ can be

872: recovered as the eigenvector of some operator such as ${dy\over

873: ds}$.

874: 

875: 

876: 

877: \subsection{ Representation on Gaudin's determinants.}

878: 

879: It is well known that the Bethe scalar products \cite{Gaudin} can be

880: expressed using a quotient of two determinants. Here, we construct a

881: representation of the T.L. algebra acting on these quotients. We

882: split the variables $z_i$ into $A=\{z_1,...,z_{n\over 2}\}$ and

883: $B=\{z_{{n\over 2}+1},...,z_{n}\}$. We also introduce $p$ a square

884: root of $q$, $p^2=q$. We define the polynomial $F'_{\omega}$:

885: 

886: \begin{eqnarray}

887: F'_{\omega}(z_1,..,z_n)={{\left|

888: (p^2z_i-p^{-2}z_j)^{-1}(pz_i-p^{-1}z_j)^{-1}\right|}\over {\left|

889: (pz_i-p^{-1}z_j)^{-1}\right|}}\Delta_{n}^q(z_1,...,z_n),\ {\rm\

890: with\ } i\in A,\ j\in B.

891:  \label{Gaudindeterminant}

892: \end{eqnarray}

893: 

894: The first factor is the ratio of the Gaudin determinant with the

895: Cauchy determinant \cite{Gaudin}. It is also related to the domain

896: wall boundary condition partition function \cite{korepin} of a six

897: vertex model with weights: $a=qx-q^{-1}y,b=px-p^{-1}y,c=\sqrt

898: {xy}(p-p^{-1})$ \footnote{Notice that for this six vertex model,

899: ${a^2+b^2-c^2\over ab}=p+{ p^{-1}}\ne\tau$.}.

900: 

901: 

902: The second factor $\Delta_n$ (\ref{deltadef}) transforms the scalar

903: product into a polynomial. This factor has an innocuous effect on

904: the T.L. algebra since:

905: 

906: \begin{eqnarray}

907: \Delta_{n}^q(z_1,...,z_n)t_i=\tilde t_i\Delta_{n}^q(z_1,...,z_n),

908:  \label{innocuous}

909: \end{eqnarray}

910: where $\tilde t_i$ is obtained from $t_i$ (\ref{polyactionTL}) by

911: the substitution $q\to -1/q$ which preserves the relations

912: (\ref{braid},\ref{hecke}), but exchanges $e_i$ with $e_i-\tau$.

913: 

914: The ratio of the two determinants being symmetrical in the two sets

915: of variables $A$ and $B$, $F'_{\omega}$ is annihilated by all the

916: $e_i$ with $i\ne{n\over 2}$.

917: 

918: \smallskip

919: 

920: To show that the action of the T.L. algebra (\ref{polynomeTLei}) on

921: $F'_\omega$ produces an irreducible module, we proceed as in

922: \ref{moduleM'}. Consider the space $\mathcal{M}_n$ of homogenous

923: polynomials in $n$ variables, and of the minimal total degree,

924: obeying the property:

925: \begin{eqnarray}

926:  {\rm(P'):}&&\ \ \bar\psi(z_{i_1}=q^{a_1}z,z_{i_2}=q^{a_2}z,z_{i_3}=q^{a_3}z)=0,

927:  \ \ {\rm if}\ i_1,i_2,i_3,\ {\rm are\ cyclically\

928:  ordered,}\cr && {\rm and\ for:}\ \

929:  (a_1,a_2,a_3)=(-1,0,1),(-1,1,0),(-2,0,2),(0,-1,1).

930:  \label{propiete2}

931: \end{eqnarray}

932: Note that these triplets are stable under the cyclic permutation,

933: $(a_1,a_2,a_3)\to(a_3-2,a_2+1,a_1+1)$, and the transpositions,

934: $(a_i,a_{i+1})\to (a_{i+1},a_{i}),$ whenever $|a_{i+1}-a_{i}|=1$.

935: 

936: >From the cyclic invariance, we deduce that this space is preserved

937: under the action of $\sigma$ (\ref{sigmacycle1}) if we take $s=q^3$.

938: 

939: By applying  the property (P') to $z_1,z_2,z_i$ with

940: $(a_1,a_2,a_3)=(-1,1,0)$ and $(-2,0,2)$, we can define a projection

941: (\ref{definitE'}) from $\mathcal{M}_n\to \mathcal{M}_{n-2}$. The

942: polynomial $\phi(z,z_i)$ is now a product of two factors

943: $\phi(z,z_i)=\prod_{i=3}^{n}(qz-z_i)(q^4z-z_i)$. Arguing as in

944: \ref{moduleM'}, we see that this projection satisfies the properties

945: (\ref{condition1E'}).

946: 

947: This space is stable under the action of the generators $e_i$. The

948: proof is similar to the one given in \ref{moduleM'}  and requires

949: the stability of the triplets $(a_1,a_2,a_3)$ under the

950: transpositions. The minimal degree is now $n({n\over

951: 2}-1)=2|\lambda|$ with $|\lambda|$ given by (\ref{partition}).

952: 

953: \smallskip

954: 

955: Let us show that $F'_{\omega}$ (\ref{Gaudindeterminant}) satisfies

956: the property (P') (\ref{propiete2}). We consider $(i_1,i_2,i_3)$ and

957: $(a_1,a_2,a_3)$. If the variables $z_l,z_m$  with $l<m$,

958: corresponding to two $a_i$ which differ by $2$, belong to the same

959: set $A$ or $B$, $F'_\omega(z_m=q^2z_l)=0$ due to the factor

960: $\Delta_n$. Otherwise, two variables $z_l=z\in A$ and $z_m=q^2z\in

961: B$ differ by a factor $q^2$. By isolating the contribution of the

962: pole $(p^2z_l-p^{-2}z_m)$ in the Gaudin determinant, we factorize a

963: term $\prod_i(qz-z_i)$ coming from the Cauchy denominator, and this

964: enables to conclude that

965: $F'_\omega(z_{i_1}=q^{a_1}z,z_{i_2}=q^{a_2}z,z_{i_3}=q^{a_3}z)=0$ in

966: all the other cases.

967: 

968: 

969: 

970: \smallskip

971: 

972: Arguing as in \ref{moduleM'} we conclude that $\mathcal{M}_n$ is an

973: irreducible module over the affine T.L. algebra and that it

974: coincides with the module obtained upon acting with the generators

975: on $F'_\omega$.

976: 

977: \smallskip

978: 

979: We verify again that the eigenvalue of the central operators $y$

980: (\ref{valpropY}) is given by (\ref{valpropy}). Now, $s=q^3$ instead

981: of $q^6$ in \ref{moduleM'}, but the degree $2\lambda$

982: (\ref{partition}) of the highest polynomial in $\mathcal{M}_n$ is

983: doubled and $c$ keeps the same value $c=q^{3(1-{n\over 2})}$.

984: 

985: In the nonsemisimple case $q^2+q+1=0$, using the result of

986: \cite{stroganov} we have $F'_\pi= s_{\lambda}F_\pi$, and therefore,

987: $\sum_\pi F'_{\pi}= s_{\lambda}^2$.

988: 

989: 

990: \section{Conclusion.}

991: 

992: Let us conclude with a few comments and questions.

993: 

994: On the mathematical side, this work provides a unification ground

995: around the conjectures relating the eigenvector components of a loop

996: model transfer matrix, the six vertex model domain wall boundary

997: condition partition function and other mathematical objects. It

998: opens the possibility to deform the polynomials underlying these

999: conjectures by presenting them from the algebra representation point

1000: of view. We believe that these conjectures are related to

1001: incompressibility, and we hope to return to this point in a future

1002: publication.

1003: 

1004: >From a technical point of view, it would be interesting to repeat

1005: the Jones construction of \ref{identify} on the polynomials

1006: directly. This would allow to recover in a direct way the product

1007: structure which they carry since they are associated to elements of

1008: the T.L. algebra.

1009: 

1010: The precise correspondence between the  polynomial obtained here

1011: and the Macdonald polynomials needs to be clarified.

1012: 

1013: Finally, do the deformed wave functions considered here have

1014: anything to do with physics? At this moment, we have no answer to

1015: this question. A step towards a physical interpretation would be to

1016: identify a scalar product and a Hermitian Hamiltonian for which

1017: these wave functions are the ground states. This could also be

1018: useful to access to the excited states which play an important role

1019: in the Quantum Hall Effect.

1020: 

1021: 

1022: \subsection{Acknowledgements}

1023: 

1024: I wish to thank Philippe di Francesco for generously explaining me

1025: his works and for discussions.

1026: 

1027: I am greatly indebted to Kirone Mallick, Gregoire Misguich and

1028: particularly Bertrand Duplantier for their help during the course of

1029: this work.

1030: 

1031: 

1032: 

1033: 

1034: 

1035: 

1036: 

1037: 

1038: 

1039: 

1040: \bigskip

1041: 

1042: 

1043: \appendix

1044: \section{Word representation.\label{wordrepresentation}}

1045: 

1046: \subsection{ Reduced words.}

1047: 

1048: The module $\mathcal{H}_n$ is obtained by acting with the T.L.

1049: generators of $\mathcal {A}_n$ on the lowest state

1050: $\alpha=e_1e_3...e_{n-1}$. Using the relations (\ref{T.L.}), we

1051: obtain a basis of $\mathcal{H}_n$ given by reduced words $\pi$:

1052: \begin{eqnarray}

1053: \pi=(e_{a_{n-1}}e_{a_{n-1}+1}..e_{n-1})...(e_{a_{2p+1}}e_{a_{2p+1}+1}..e_{2p+1})...

1054: (e_{a_3}e_{a_3+1}..e_3)e_1,\label{word}

1055: \end{eqnarray}

1056: with, $a_{2p+1}\le 2p+1$, and $1<a_3<..< a_{2p+1}<..<a_{n-1} $. So,

1057: a word is fully characterized by the sequence $(a_{2p+1})$.

1058: 

1059: On reduced words there is a natural order relation: $\pi>\pi'$ if

1060: $\pi$ is written $b{\pi'}$ with b a monomial. One has $\pi\ge\pi'$

1061: if $a_{2p+1}\le a'_{2p+1}$ for all $p$.

1062: \smallskip

1063: 

1064: Another way to represent a reduced word is in terms of paths. Let

1065: $m_i$ be the number of times the generator $e_i$ appears in the

1066: reduced expression of $\pi$. One has $m_{2i}=m_{2i-1} {\rm \ or}\

1067: m_{2i-1}-1$ and $m_{2i+1}=m_{2i} {\rm \ or}\ m_{2i}+1$. We define

1068: $h_{2i}=2m_{2i}-1$, $h_{2i-1}=2m_{2i-1}-2$ and $h_0=h_n=0$ by

1069: convention. We can describe the words $\pi$ by the paths $\pi=[h_i]$

1070: where $h_0=h_{n}=0$, $h_i\ge 0$ and $h_{i+1}-h_{i}=\pm 1$. Using the

1071: path representation, one has $\pi\ge\pi'$, if $[h_i]\ge[h'_i]\

1072: \forall i.$

1073: 

1074: If $\pi'$ is a non reduced word, by reducing it, one decreases the

1075: number of times the generator $e_i$ appears in its expression. We

1076: thus see that the order relation can be presented in a weaker form

1077: valid for non reduced words: If $\pi'$ is a word, not necessarily

1078: reduced and $\pi$ is a reduced word, $\pi>\pi'$ if $\pi'$ can be

1079: obtained by erasing letters  $e_k$ from the (reduced) expression of

1080: $\pi$.

1081: 

1082: \smallskip

1083: Finally, there is way to characterize this representation in terms

1084: of link patterns. It is convenient to dispose the $n$ points

1085: cyclically around a circle. A link pattern is obtained by pairing

1086: all the points in the set $\{1,2,...,n\}$:

1087: $\pi=\{[i_1,i_2],[i_3,i_4],...,[i_{n-1},i_n]\}$, in such a way that

1088: two links never cross. In practise, if $[i,j]$ is a link, then the

1089: other links $[k,l]$ are either inside, or outside the interval

1090: $[i,j]$. The action of $e_{i,i+1}$ is given by:

1091: $e_{i,i+1}\{[i,i+1],...,[i_{n-1},i_n]\}=\tau

1092: \{[i,i+1],...,[i_{n-1},i_n]\}$, and

1093: $e_{i,i+1}\{[i,j],[i+1,k]...,[i_{n-1},i_n]\}=\{[i,i+1],[j,k],...\}$.

1094: In this representation, $\alpha=\{[1,2],[3,4],...,[n-1,n]\}$, and

1095: $\omega=\{[1,n],[2,n-1],...,[{n\over 2}-1,{n\over 2}+1]\}$.

1096: 

1097: These representations are illustrated in figure

1098: \ref{fig:wordrepresentation}.

1099: 

1100: \begin{figure}

1101: 

1102: \begin{center}

1103: 

1104: \includegraphics[width=6cm]{zigzag.eps}

1105: 

1106: \caption[99]{

1107: 

1108: The three different ways to represent a word illustrated in the case

1109: of $\mathcal{H}_6$.}

1110: 

1111: \label{fig:wordrepresentation}

1112: 

1113: \end{center}

1114: 

1115: \end{figure}

1116: 

1117: 

1118: 

1119: 

1120: \subsection{ Identifying $\mathcal{H}_n$ with $\mathcal{A}_{n\over

1121: 2}.$\label{identify}}

1122: 

1123: The link pattern representation allows to identify in a natural way

1124: $\mathcal{H}_n$ with $\mathcal {A}_{n\over 2}.$ If we split

1125: $\{1,2,...,n\}$ into two subsets :$ \{1,2,...,{n\over 2}\}$ and $

1126: \{{n\over 2},...,n\}$, the product $\pi*\pi'$ is defined on the link

1127: patterns by ``stacking'' the two link patterns and identifying the

1128: last ${n\over 2}$ points of $\pi$ with the first ${n\over 2}$ points

1129: of $\pi'$ through $i\equiv n+1-i$. The link pattern $\pi*\pi'$ is

1130: obtained by removing the loops which appear in this concatenating

1131: operation by giving them a weight $\tau$.

1132: 

1133: Another identification can be achieved on paths by folding a path of

1134: length $n$ into a loop of length ${n\over 2}$. In this way, we

1135: realize $\mathcal {A}_{n\over 2}$ as the algebra of double paths

1136: acting on Bratteli diagrams

1137: \cite{Pasbratel}\cite{Pasthese}\cite{Jones}.

1138: 

1139: In this identification, $\mathcal {A}_{n\over 2}$ is a bimodule over

1140: itself. The first ${n\over 2}-1$ generators $e_i\in \mathcal

1141: {A}_{n}$ are identified with the generators of $\mathcal {A}_{n\over

1142: 2}$ acting to the left, while the last ${n\over 2}-1$ generators are

1143: identified with $e_{n+1-i}$ acting to the right.

1144: 

1145: The state $\omega$ is the identity in $\mathcal {A}_{n\over 2}$, and

1146: the trace in $\mathcal {A}_{n\over 2}$ coincides with the scalar

1147: product with $\omega$ in $\mathcal {A}_{n}$:

1148: \begin{eqnarray}

1149: {\rm t r} (x)=\tau^{-{n\over 2}}\langle \omega |x\rangle .

1150: \label{identifytrace}

1151: \end{eqnarray}

1152: 

1153: The projection: $E_{{n\over 2}}=\sigma^{-{n\over

1154: 2}+1}E\sigma^{{n\over 2}-1},$ with $E$ given by (\ref{project-2})

1155: can be reinterpreted as a conditional expectation value

1156: \cite{Jones}, $E_{n\over 2}:\ \mathcal {A}_{n\over 2}\to \mathcal

1157: {A}_{{n\over 2}-1}$. Jones construction enables then to construct

1158: $e_{n\over 2}\in \mathcal {A}_{{n\over 2}+1}$ algebraically from the

1159: knowledge of  $E_{n\over 2}$.

1160: 

1161: 

1162: 

1163: \bigskip \bigskip

1164: \subsection{ Triangularity of $y_m$.\label{triangularity}}

1165: 

1166: Let us show that the affine generators

1167: $y_{m+1}=t^{-1}_mt^{-1}_{m-1}...t_1^{-2}...t^{-1}_m$ are triangular

1168: in the word representation. It is obvious for $y_1=1$ and

1169: $y_2=t_1^{-2}$ since $e_1$ is triangular. We proceed by recursion

1170: and assume that $y_{k}$ are triangular for $k<m+1$. Using these

1171: hypotheses, we show that $y_{m+1}$ is also triangular.

1172: 

1173: First we show that $y_{m+1}$ acts diagonally on $\alpha$. To study

1174: the action of $y_{m+1}$ on $\alpha$, we distinguish the two cases

1175: $m$ odd or even. If m is odd, then:

1176: \begin{eqnarray}

1177: y_{m+1}\alpha=t^{-1}_my_{m}t^{-1}_me_m...=-qt^{-1}_my_{m}e_m...=-\lambda_{m}qt^{-1}_me_m...=q^2\lambda_{m}\alpha,

1178: \end{eqnarray}

1179: where $\lambda_m$ is the eigenvalue of $y_{m}$ on $\alpha$. If $m$

1180: is even, we make use of the fact that

1181: $t_{m-1}^{-1}t^{-1}_me_{m-1}={1\over q}e_me_{m-1}$ and the same

1182: relation with the indices $m$ and $m-1$ exchanged to obtain:

1183: \begin{eqnarray}

1184: y_{m+1}\alpha&=&t^{-1}_mt^{-1}_{m-1}y_{m-1}t^{-1}_{m-1}t^{-1}_me_m...={1\over

1185: q}t^{-1}_my_{m-1}e_me_{m-1}...\cr &=&{1\over

1186: q}\lambda_{m-1}t^{-1}_mt^{-1}_{m-1}e_me_{m-1}...={1\over

1187: q^2}\lambda_{m-1}\alpha.

1188: \end{eqnarray}

1189: We deduce that $\alpha$ is an eigenstate of $y_m$ with the

1190: eigenvalue $\lambda_{m}$ obeying the recursion relations

1191: $\lambda_{2m}=q^2\lambda_{2m-1}$, $\lambda_{2m+1}={1\over

1192: q^2}\lambda_{2m-1}$. Together with the fact that $\lambda_1=1$, we

1193: deduce (\ref{valpropY}).

1194: 

1195: \bigskip

1196: 

1197: 

1198: 

1199: To show that $y_{m+1}$ is triangular on words $\ne \alpha$. We

1200: proceed by recursion and assume that $y_{m+1}$ acts in a triangular

1201: way on words  $<\pi$ and show that the property is also true for

1202: $\pi$.

1203: 

1204: Let us consider the action of $y_{m+1}$ on a reduced word $\pi\ne

1205: \alpha$. This word can be put under  the form $\pi=e_i\pi'$ where

1206: $\pi'<\pi$. We consider the three cases, $i\ne m,m+1$, $i=m$,

1207: $i=m+1$. In the third case, either the word can be written in the

1208: form $e_{m+1}e_m\pi'$ with $\pi'$ reduced, or it can be written

1209: $e_p\pi'$ with $p<m$. The second possibility reduces to the first

1210: case and we need only consider the first possibility.

1211: 

1212: We observe that $y_{m+1}$ commutes with $e_i$:

1213: $y_{m+1}e_i=e_iy_{m+1}$ if $i>m+1$ or if $i<m$. It is obvious if

1214: $i>m+1$ and follows from the braid relations if $i<m$. In the three

1215: cases we can thus write:

1216: 

1217: 

1218: 

1219: \begin{eqnarray}

1220: y_{m+1}e_i\pi'&=&e_i(y_{m+1}\pi')\ \ {\rm for}\ i\ne m,m+1,\cr

1221: y_{m+1}e_m\pi'&=&t^{-1}_my_{m}t^{-1}_{m}e_m\pi'=-qt^{-1}_m(y_{m}e_m\pi'),\cr

1222: y_{m+1}e_{m+1}e_m\pi'&=&t^{-1}_my_{m}t^{-1}_{m}e_{m+1}e_m\pi'=t^{-1}_m(y_{m}e_m\pi'

1223:  +{1\over q}y_me_{m+1}e_m\pi').

1224: \end{eqnarray}

1225: It follows from the hypothesis that the terms in brackets are less

1226: than $\pi$. In the first case because $\pi'<\pi$, and in the two

1227: others because $y_m$ is assumed to be triangular.

1228: 

1229: To conclude that $y_{m+1}$ is triangular, we must show that the

1230: action of $e_i$ in the first case and  $e_m$ in the two other cases

1231: preserves the triangularity : If $e_i\pi$ is a reduced word and

1232: $\pi'\leq \pi$, then, $e_i\pi'\leq e_i\pi$. If $e_m\pi$ is a reduced

1233: word and $\pi'\leq e_m\pi$, then $e_m\pi'\leq e_m\pi$. Finally, if

1234: $e_{m+1}e_m\pi$ is a reduced word and $\pi'\leq e_{m+1}e_m\pi$, then

1235: $e_m\pi'\leq e_{m+1}e_m\pi$. These properties follow from the weak

1236: form of the order relation. This concludes the proof of

1237: triangularity of $y_{m+1}$.

1238: 

1239: \bigskip

1240: \subsection{  Action of $\sigma$ on words.\label{actionofsigmaonwords}}

1241: 

1242: The action of $\sigma=-q^{{n\over 2}-2}t_{n-1}^{-1}...t_1^{-1}$ on

1243: words can be computed similarly. First, using the braid relation

1244: (\ref{braid}), one sees that $\sigma e_i=e_{i-1}\sigma$ for $i>1$.

1245: To fully characterize its action, we must compute $(\sigma \alpha)$.

1246: Using $t_1^{-1}e_1=-qe_1$ and $t_{m+1}^{-1}t^{-1}_me_{m+1}={1\over

1247: q}e_me_{m+1}$  we obtain:

1248: \begin{eqnarray}

1249: \sigma \alpha= \prod_{i=1}^{{n\over 2}-1}e_{2i}\alpha.

1250: \end{eqnarray}

1251: Thus, $(\sigma \alpha)$ can be characterized by the property:

1252: \begin{eqnarray}

1253: e_{2i}(\sigma \alpha)=\tau (\sigma \alpha),

1254: \end{eqnarray}

1255: for $1\le i\le {n\over 2}$.  $(\sigma\alpha)$ can then be used as a

1256: lowest state to construct a canonical basis by acting on it with the

1257: generators $e_2,...,e_{n}$.

1258: 

1259: \bigskip

1260: 

1261: \section{Explicit construction of $\mathcal{M}_4$.\label{H4}}

1262: 

1263: Let us construct $\mathcal{M}_4$ the dual of $\mathcal{H}_4$. The

1264: basis of $\mathcal{H}_4$ is given by the words $e_1e_3=\alpha,\

1265: e_2e_1e_3=\omega$. So we search for a vector $\Psi$ of the form:

1266: \begin{eqnarray}

1267: \Psi=F_{\alpha}(z_1,..,z_4)\alpha+F_{\omega}(z_1,..,z_4)\omega,\label{Psi4}

1268: \end{eqnarray}

1269: where $F_{\alpha},F_{\omega}$ are polynomials of degree $(1,1)$ in

1270: the variables $z_i$.

1271:  The action of the $T.L.$ affine algebra is given by the matrices:

1272: \begin{eqnarray}

1273: e_1=e_3=\pmatrix{\tau&1\cr 0&0}, \ e_2=e_4=\pmatrix{0&0\cr 1&\tau},\

1274: \sigma =\pmatrix{0&1\cr 1&0}. \label{TL4}

1275: \end{eqnarray}

1276: 

1277: We can obtain the dual representation by acting with the generators

1278: on  $F_{\omega}\equiv \pmatrix{0,1}$ annihilated by $\bar

1279: e_1=e_1,\bar e_3=e_3$. The minimum degree polynomial annihilated by

1280: $ e_1, e_3$ is given by:

1281: \begin{eqnarray}

1282: F_{\omega}=(qz_1-q^{-1}z_{2})(qz_3-q^{-1}z_{4}). \label{F2-4}

1283: \end{eqnarray}

1284: 

1285: Let us take $\bar\sigma=\sigma$ of the form:

1286: \begin{eqnarray}

1287: F(z_1,z_2,z_3,z_4)\sigma =c F(z_2,z_3,z_4,sz_1). \label{C-4}

1288: \end{eqnarray}

1289: We obtain two different expression for $F_{\alpha}\equiv(1,0)$ which

1290: we must equate. One  results from the cyclic property:

1291: $F_{\alpha}=F_{\omega}\sigma$, the other given by:

1292: $F_{\alpha}=F_{\omega}(e_2-\tau)$.

1293: 

1294: We get the equation:

1295: \begin{eqnarray}

1296: {(qz_1-q^{-1}z_2)(qz_3-q^{-1}z_4)-(qz_1-q^{-1}z_3)(qz_2-q^{-1}z_4)\over

1297: z_2-z_3}=c (qz_4-q^{-1}sz_1), \label{miracle-4}

1298: \end{eqnarray}

1299: which determines $s=q^6$, $c=q^{-3}$, and:

1300: 

1301: \begin{eqnarray}

1302: F_{\alpha}=(qz_2-q^{-1}z_{3})(q^{-2}z_4-q^{2}z_{1}). \label{F2-4}

1303: \end{eqnarray}

1304: 

1305: 

1306: \bigskip

1307: 

1308: \section{ Module $F_{\omega}$.\label{module}} Let us define a T.L.

1309: module $M$ defined in terms of a state $F_{\omega}$ satisfying

1310: $F_{\omega }e_i=0$ for $i\ne {n\over 2}$. The module is obtained by

1311: acting with the T.L. generators and reducing words using the T.L.

1312: relations (\ref{T.L.}). In this module, a canonical basis is:

1313: \begin{eqnarray}

1314: \bar \psi=F_{\omega}(e_{n\over 2}e_{{n\over 2}-1}...e_{a_{n\over

1315: 2}+1}e_{a_{n\over2}})...

1316: (e_{p}e_{{p}-1}...e_{a_{p}})...(e_{n-1}...e_{a_{n-1}}),\label{word1}

1317: \end{eqnarray}

1318: where the $p$ take the all the values between ${n\over 2}$ and $n-1$

1319: and the $a_{p}$ are restricted by the conditions: $a_{p}\le p+1$,

1320: $a_{n\over 2}<a_{{n\over 2}+1}..< a_{p}<..<a_{n-1}$. The convention

1321: is that if $a_p=p+1$, the sequence $(e_{p}...e_{a_{p}})$ is empty. A

1322: word $\bar \psi$ is fully characterized by the sequence $(a_{p})$.

1323: The word can also be associated to the Young diagram

1324: $[\mu_{p+1-{n\over 2}}]=[p-a_p+1]$.

1325: 

1326: There is a reflection symmetry, $i\to n-i$, and an alternative

1327: description of the module in terms of reflected words:

1328: \begin{eqnarray}

1329: \bar \psi=F_{\omega}(e_{n\over 2}...e_{b_{n\over 2}-1}e_{b_{n\over

1330: 2}})...(e_{p}...e_{b_{p}-1}e_{b_{p}})...(e_{1}...e_{b_{1}}),\label{word11}

1331: \end{eqnarray}

1332: $1\le p\le {n\over 2}$, $b_p\ge p-1$, $b_{n\over 2}>...>b_1$. It is

1333: associated to the dual Young diagram $[\mu'_{{n\over

1334: 2}-p+1}]=[b_p-p+1]$

1335: 

1336: A similar order relation as defined earlier holds for reduced words,

1337: $\bar\psi'<\bar\psi$ if $\bar\psi$ can be written

1338: $\bar\psi=\bar\psi'a$. For non reduced words $\bar\psi'$, it is

1339: sufficient that $\bar\psi'$ can be obtained by erasing letters $e_k$

1340: from the (reduced) expression of $\bar\psi$.

1341: 

1342: \smallskip

1343:  In general, the module $F_{\omega}$ is reducible, it will

1344: be irreducible if $F_{\omega}$ satisfies the Fock condition:

1345: \begin{eqnarray}

1346: F_{\omega}(1+\sum_{m=0}^{{n\over 2}-1} q^{m+1}t_{{n\over

1347: 2}}...t_{{n\over 2}-m})=0.\label{fock}

1348: \end{eqnarray}

1349: 

1350: In this case, The only allowed words $\bar\psi$ (\ref{word1}) can be

1351: associated to their complementary $\pi_{\psi}$ in such a way that

1352: one can write without reducing the expression:

1353: \begin{eqnarray}

1354: \psi\pi_{\psi} =\omega.\label{motcomplement}

1355: \end{eqnarray}

1356: Thus, we get the supplementary constraint $a_{p}>2p+1-{n}, \

1357: b_p<2p-1$.

1358: 

1359: 

1360: 

1361: \subsection{ Reducing the Hecke Module to its T.L. form.\label{formedehecke}}

1362: 

1363: Let us consider a module $M'$ over the Hecke algebra defined by

1364: acting with the Hecke algebra generators satisfying (\ref{hecke1})

1365: on the state $F_{\omega}$ satisfying $F_{\omega} e_i=0$ for $i\ne

1366: {n\over 2}$ . We want to show that the Hecke algebra acts as a T.L.

1367: algebra on this module. For this, we first show that the Hecke

1368: relations (\ref{hecke1}) are sufficient to reduce the word basis of

1369: $M'$ to be of the T.L. form (\ref{word1}). Thus, $M'$ and $M$ can be

1370: identified as vector spaces. From this, we will deduce that $M'=M$

1371: as modules. In other words, the projectors

1372: $U^-_{i,i+1}=e_ie_{i+1}e_i-e_i$ are null in $M'$.

1373: 

1374: \smallskip

1375: Let us assume that it is not true. Since all the basis elements of

1376: $M'$ are obtained upon acting on $F_{\omega}$ with letters $e_k$,

1377: there is a basis element $\bar\psi e_i$ which cannot be expressed as

1378: a linear combination of words of the form (\ref{word1}) although

1379: $\bar\psi$ is of the form (\ref{word1}). Among all the $\bar\psi$

1380: which verify this property, we can take the smallest possible for

1381: the order relation, so that that $\bar\psi'e_i$ is of the form

1382: (\ref{word1}) when $\bar\psi'<\bar\psi$. We show that this leads to

1383: a contradiction.

1384: 

1385: Let us consider the word $\bar\psi e_i$. It is a word of the form

1386: (\ref{word1}) in the three following cases. When $\bar\psi e_i$ is a

1387: reduced word $>\bar\psi$, for $i=a_p-1$ if $a_p-1>a_{p-1}$. When

1388: $\bar\psi e_i=\tau \bar\psi$ when $i=a_p$ and $a_p>a_{p+1}-1$. When

1389: $\bar\psi e_i=0$ if $i<a_{n\over 2}-1$ or $i>b_{n\over 2}+1$.

1390: 

1391: The two remaining cases to consider are: First, when $a_p< i<

1392: a_{p+1}-1$ for some $p$. Second, when $a_p<i \le a_p+k$ if

1393: $a_{p+k}=a_p+k$ with $k\ge 1$. The second case can be studied

1394: similarly to the first one using the reflection symmetry $i\to n-i$

1395: and corresponds to $b_{p'}>i>b_{p'-1}+1$.

1396: 

1397: 

1398: In the first case, $\bar\psi

1399: e_i=\bar\psi'(e_p...e_{a_p+1}e_{a_p})e_i(e_{p+1}...e_{a_{p+1}})...$,

1400: and using the relation (\ref{hecke1}), we see that:

1401: \begin{eqnarray}

1402: e_p...e_{a_p+1}e_{a_p}e_i      = e_{i-1}e_p ...e_{a_p+1}e_{a_p}+

1403: e_p... e_{i+1}(e_i-e_{i-1}) e_{i-2}...e_{a_p+1}e_{a_p}.\label{sueur}

1404: \end{eqnarray}

1405: The second term is $<\bar\psi$ and therefore of the T.L. form by the

1406: recursion hypothesis. The first term can be eliminated by repeating

1407: this relation $p-{n\over 2}$ times to push $e_{i}$ and then

1408: $e_{i-1},...,e_{i+{n\over 2}-p}$ to the left of the word. The last

1409: application of the relation gives a term $F_{\omega} e_{i+{n\over

1410: 2}-p-1}=0$ since $i+{n\over 2}-p-1<{n\over 2}$.

1411: 

1412: This exhaust all the possibilities and $\bar\psi e_i$ can always be

1413: expressed as a linear combination of reduced T.L. words

1414: (\ref{word1}) in contradiction with the hypothesis. Therefore, the

1415: word basis of $M'$ coincides with the word basis (\ref{word1}).

1416: 

1417: \smallskip

1418: To conclude that $M'=M$, let us consider the projectors

1419: $U^-_{i,i+1}=e_ie_{i+1}e_i-e_i$, and the space $M''\subset M'$

1420: annihilated by all the $U^-_{i,i+1}$. The space $M''$ defines a

1421: module for the T.L. algebra. Since $F_\omega\in M''$, this module

1422: can be identified with $M$. Therefore, $M$ is a subspace of $M'$

1423: with the same dimension, and thus, $M=M'$.

1424: 

1425: \bigskip

1426: 

1427: \section{Yang-Baxter Equation and Polynomials:}

1428: 

1429: 

1430: \subsection{Polynomial representation of the Hecke generators

1431: \label{poly-hecke}}

1432: 

1433: In this section, we derive the expression of the Hecke generators

1434: $\bar t_i$ (\ref{polyactionTL}) from the Yang-Baxter equation.

1435: 

1436: The Yang-Baxter algebra \cite{Gaudin} (also called $RLL=LLR$

1437: relation) can be expressed as:

1438: \begin{eqnarray}

1439:  R_{12}(z_1,z_2)L_1(z_1)L_2(z_2)=L_2(z_2)L_1(z_1)R_{12}(z_1,z_2),

1440: \label{yang-baxter1} \end{eqnarray} where $R_{12}(z_1,z_2)$ is a

1441: solution of the Yang-Baxter equation:

1442: \begin{eqnarray}

1443:  R_{12}(z_1,z_2)R_{13}(z_1,z_3)R_{23}(z_2,z_3)=R_{23}(z_2,z_3)R_{13}(z_1,z_3)R_{12}(z_1,z_2).

1444: \label{yang-baxter0} \end{eqnarray}

1445: 

1446: If we assume that $R_{12}(z_1,z_2)=Y_{12}(z_1,z_2)P_{12}$ where

1447: $P_{12}$ acts in the natural way on the spin indices,

1448: $(P_{12}t_{13}=t_{23}P_{12}),$  but commutes with $z_i$,

1449: (\ref{yang-baxter1}) rewrites as:

1450: \begin{eqnarray}

1451:  Y_{12}(z_1,z_2)L_2(z_1)L_1(z_2)=L_2(z_2)L_1(z_1)Y_{12}(z_1,z_2)=L_2(z_1)L_1(z_2)k_{12},

1452: \label{yang-baxter2}

1453: \end{eqnarray}

1454: where $k_{12}$ acts to the left by permuting the variables

1455: $z_1,z_2$. If we normalize of $Y(z_1,z_2)$ so that:

1456: \begin{eqnarray}

1457: Y_{12}(z_1,z_2)Y_{12}(z_2,z_1)=1, \label{YY-yang-baxter2}

1458: \end{eqnarray}

1459: it is consistent to demand that the $Y_{ii+1}$ act as a

1460: representation of the permutation algebra on some wave function

1461: $\Psi$:

1462: \begin{eqnarray}

1463:  Y_{12}(z_1,z_2)\Psi(z_1,z_2)=\Psi(z_2,z_1)=\Psi(z_1,z_2)k_{12}.

1464: \label{Y=k}

1465: \end{eqnarray}

1466: The  $Y_{ij}$ are are called Yang's operators in \cite{Gaudin}.

1467: 

1468: A well known solution of

1469: (\ref{yang-baxter0}),(\ref{YY-yang-baxter2}) in terms of the Hecke

1470: algebra (\ref{hecke}) is:

1471: \begin{eqnarray}

1472:  Y_{12}(z)={ t_{12}-z t_{12}^{-1}\over z q -

1473:  q^{-1}},

1474: \label{yang-baxter3}

1475: \end{eqnarray}

1476: where $z={z_1\over z_2}$.

1477: 

1478: Substituting (\ref{yang-baxter3}) in (\ref{Y=k}), we can also

1479: rewrite this relation as:

1480: \begin{eqnarray}

1481:  t_{12}\Psi(z_1,z_2)=\Psi(z_1,z_2)\bar t_{12},

1482: \label{t=g}

1483: \end{eqnarray}

1484: Where $\bar t_{12}$ takes the form:

1485: \begin{eqnarray}

1486: \bar t_{12}=-q^{-1}+(1-k_{12}){qz_1-q^{-1}z_2\over z_1-z_2},

1487:  \label{polyactionTL1}

1488: \end{eqnarray}

1489: and this coincides with (\ref{polyactionTL}).

1490: 

1491: \subsection{Commutation relations of the affine generators $y_i$.\label{discussion}}

1492: 

1493: We motivate the commutation relation (\ref{affine-gener}c) from the

1494: Yang-Baxter algebra (\ref{yang-baxter1}) point of view. This also

1495: reveals a complete symmetry between the spectral parameters $z_i$

1496: and the generators $y_i$.

1497: 

1498: 

1499: Let us substitute the spectral parameters $z_i$ with the affine

1500: generators $y_i$ in $L(z_i)$, and we require that the relation

1501: (\ref{t=g}) are preserved under the action of the algebra $L_i$ on

1502: $\Psi$:

1503: \begin{eqnarray}

1504: t_{12}L_1(y_1)L_2(y_2)\Psi=L_1(y_1)L_2(y_2)\Psi\bar t_{12},

1505:  \label{consistance}

1506: \end{eqnarray}

1507: assuming that (\ref{t=g}) holds for $\Psi$.

1508: 

1509: To avoid cumbersome expressions, we use from here the transposed

1510: notation $\bar a X$ for $X \bar a$. We must then transpose back the

1511: final algebraic relations we deduce. In the transposed notations

1512: (\ref{consistance}) is equivalent to:

1513: \begin{eqnarray}

1514:  (t_{12}-\bar t_{12})L_1(y_1)L_2(y_2)=0,

1515:  \label{consistance1}

1516: \end{eqnarray}

1517: under the hypothesis that $t_{12}=\bar t_{12}$ to the right of any

1518: expression. Let us for the moment assume that $\bar t_{12}$ commutes

1519: with the symmetrical expressions in $y_1,y_2$.

1520: 

1521: After substituting the expression of $L_i(y_i)$ deduced from

1522: (\ref{yang-baxter3}):

1523: \begin{eqnarray}

1524:  L_1(y_1)=(y t_{10}-y_1 t_{10}^{-1})P_{01},

1525:  \label{expressionLi}

1526: \end{eqnarray}

1527: the term proportional to $y^0$ requires that $\bar t_{12}$ commutes

1528: with $y_1 y_2$, while the term proportional to $y$ imposes that:

1529: \begin{eqnarray}

1530:  (t_{12}-\bar t_{12})(y_2t_{01}t_{12}^{-1}+y_1t_{01}^{-1}t_{12})=0,

1531:  \label{consistance2}

1532: \end{eqnarray}

1533: under the hypothesis that $t_{01}=\bar t_{12}$ to the right of any

1534: expression. This gives:

1535: \begin{eqnarray}

1536:  y_2\bar t_{12}+(q-q^{-1})y_1-\bar t_{12}y_1&=&0,\cr

1537:  y_1-\bar t_{12}y_2\bar t_{12}&=&0,

1538:  \label{consistance3}

1539: \end{eqnarray}

1540: which is equivalent to $y_2\bar t_{12}=\bar t_{12}^{-1}y_1$ and

1541: implies in particular that $\bar t_{12}$ commutes with the

1542: symmetrical expressions in $y_1,y_2$. After transposition, it yields

1543: (\ref{affine-gener}c) back.

1544: 

1545: \smallskip

1546: 

1547: Alternatively, we can substitute $z_i$ for $y_i$ in

1548: (\ref{affine-gener}c) and verify that the relation is obeyed when we

1549: use the expression (\ref{polyactionTL1}) of $\bar t_i$.

1550: 

1551: \subsection{\bf Eigenvalues of the  $y_j$ in the polynomial

1552: case.\label{diag-yi}}

1553: 

1554: We show that the operators $y_j$ defined with the polynomial

1555: representation \ref{Polynomial representations of the Hecke algebra}

1556: are triangular matrices. Let us recall the expression of $y_i$:

1557: \begin{eqnarray}

1558: y_i&=&x_{ii+1}x_{ii+2}...x_{in-1}s_ix_{i1}...x_{ii-1}\label{yang-rep00}

1559: \end{eqnarray}

1560: where the operator $x_{i,j}$ takes the form for $i<j$:

1561: \begin{eqnarray}

1562: x_{ij}=-q^{-1}+(q-q^{-1})(1-k_{ij}){z_j\over

1563: z_i-z_j},\label{polyactionX}

1564: \end{eqnarray}

1565: and the operators $s_i$ act as:

1566: \begin{eqnarray}

1567:  P(z_1,..,z_i..,z_n)s_{i}&=& c

1568: P(z_1..,sz_i,...,z_n).\label{polyaction-si}

1569: \end{eqnarray}

1570: 

1571: $x_{12}$ commutes with $z_1z_2$ and with $z_k$ for $k\ne 1,2$. It

1572: acts in the following way for on the monomials $z_1^m,\ z_2^m $:

1573: \begin{eqnarray}

1574: z_1^mx_{12}&=&-q^{-1}z_1^m+(q-q^{-1})(z_1^{m-1}z_2+z_1^{m-2}z_2^2+...+z_2^m)\cr

1575: z_2^mx_{12}&=&-qz_2^m-(q-q^{-1})(z_1^{m-1}z_2+z_1^{m-2}z_2^2+...+z_1z_2^{m-1})

1576: \label{ordreX}

1577: \end{eqnarray}

1578: Let us consider which new monomials $z^{\lambda'}$ can appear when

1579: one acts with $x_{12}$ on the monomial $z^{\lambda}$. First, all the

1580: $\lambda'_l$ for $l\ne 1,2$ are equal to $\lambda_l$. Then, if

1581: $\{\lambda'_i\lambda'_j\}\ne \{\lambda_i\lambda_j\}$ with

1582: $\{i,j\}=\{1,2\}$ and $\lambda'_j\le \lambda'_i$ , we must have

1583: $\{\lambda'_i,\lambda'_j\}= \{\lambda_i-p,\lambda_j+p\}$ for some

1584: integer $p$. Finally, if $\{\lambda'_1\lambda'_2\}=

1585: \{\lambda_1\lambda_2\}$, the only possibility is that:

1586: $(\lambda'_1,\lambda'_2)=(\lambda_2,\lambda_1)$ with

1587: $\lambda_1>\lambda_2$.

1588: 

1589: 

1590: Let us  define an order on the  monomials by saying that

1591: $z^{\lambda}$ is larger than $z^{\lambda'}$ if either $\lambda'$ is

1592: obtained from $\lambda$ by a sequence of squeezing operations

1593: $\{\lambda_i,\lambda_j\}\to \{\lambda_i-1,\lambda_j+1\}$ with

1594: ${\lambda_i>\lambda_j+1},$ or $\lambda'$ is a permutation of

1595: $\lambda$ and can be obtained from $\lambda$ by a sequence of

1596: permutations $(\lambda_i,\lambda_{i+1})\to

1597: (\lambda_{i+1},\lambda_{i})$ with $\lambda_i>\lambda_{i+1}$ . It

1598: follows from the above analysis that the action of $y_j$ on a

1599: monomial produces only monomials which are smaller with respect to

1600: this order. Thus the eigenvalues of the operators $y_j$ are given by

1601: the diagonal elements in the monomial basis.

1602: 

1603: Given the partition $\lambda=(\lambda_1,...,\lambda_n)$, the

1604: eigenvalues corresponding to the monomials associated to it are  all

1605: obtained by permutations of the multiplet:

1606: \begin{eqnarray}

1607: (y_j)=c(-q)^{1-n}(t^{\lambda_j}q^{2(j-1)}) .\label{spectre-yj}

1608: \end{eqnarray}

1609: 

1610: 

1611: 

1612: 

1613: 

1614: 

1615: 

1616: \begin{thebibliography}{99}

1617: 

1618: 

1619: \bibitem{razumov1} A.V. Razumov and Y.G. Stroganov, J.Phys. A {\bf 34}, 3185,

1620: [cond-mat/0012141].

1621: 

1622: \bibitem{razumov2} A.V. Razumov and Y.G. Stroganov, Theor.Math.Phys.

1623: {\bf 138} 333-337 (2004).

1624: 

1625: \bibitem{pierce} P.A. Pearce, V. Rittenberg, J. de Gier and B. Nienhuis,

1626: J.Phys.A {\bf 35} L661-668 (2002) [math-phys/0209017].

1627: 

1628: 

1629: \bibitem{Pdf} P.Di Francesco and P. Zinn-Justin, Electr.J.Combin.

1630: {\bf12}, R6 (2005), [math-ph/0410061].

1631: 

1632: 

1633: \bibitem{halperin} B.A. Halperin, Helv.Phys.Acta {\bf 56}, 75

1634: (1983).

1635: 

1636: \bibitem{korepin} V.E. Korepin, N.M. Bogoliubov, and A.G.Izergin,

1637: {\it Quantum Inverse Scattering Method, Correlation Functions and

1638: Algebraic Bethe Ansatz.} Cambridge University Press, New York,

1639: (1993).

1640: 

1641: 

1642: \bibitem{haldanerezai} F.D.M. Haldane and E.H. Rezayi, Phys.Rev.Lett.

1643: {\bf 60}, 956 (1988) and E{\bf 60}, 956 (1988).

1644: 

1645: 

1646: \bibitem{Gaudin} M. Gaudin, {\it La fonction d'onde de Bethe.} Masson

1647: (1983).

1648: 

1649: \bibitem{pas2} V. Pasquier, {\it Incompressible representations of the

1650: Birman Wenzl algebra.} in preparation.

1651: 

1652: \bibitem{mooreread} G. Moore and N. Read, Nuclear Phys. {\bf B360},

1653: 362 (1991).

1654: 

1655: \bibitem{birman} J. Birman and H. Wenzl, trans. A.M.S. {\bf 313}, 249

1656: (1989).

1657: 

1658: \bibitem{Pdf2} P.Di Francesco and P.Zinn-Justin, [math-ph/0412031].

1659: 

1660: \bibitem{degier} J.De Gier and B.Nienhuis, [math.AG/0410392].

1661: 

1662: \bibitem{prange} E. Prange and S. Girvin, {\it The Quantum Hall

1663: effect.} Springer-Verlag, (1987).

1664: 

1665: \bibitem{hallpasq} V. Pasquier, C.R.Physique Academie des sciences {\bf

1666: 3}, 709-15 (2002).

1667: 

1668: \bibitem{batchelor} M.T. Batchelor, J. de Gier and B.Nienhuis,

1669: J.Phys.A {\bf 34} L265-270 (2001) [cond-mat/0101385].

1670: 

1671: \bibitem{Pasthese} V. Pasquier, {\it Mod\`eles exacts invariants

1672: conformes.} Th\`ese pr\'esent\'ee \`a l'Universit\'e Paris-Sud

1673: (centre d'Orsay) (1988).

1674: 

1675: \bibitem{Jones} F.M. Goodman, P. de la Harpe, V.F.R. Jones, {\it Coxeter

1676: Graphs and Towers of Algebras.} Springer-Verlag (1989).

1677: 

1678: 

1679: \bibitem{BGHP} D. Bernard, M. Gaudin, D. Haldane and V. Pasquier,

1680: J.Phys. A {\bf 26},5219-5236 (1993), [hep-th/9301084].

1681: 

1682: \bibitem{Macdo2} I.G. Macdonald, {\it Affine Hecke Algebras and orthogonal

1683: Polynomials.} S\'eminaire Bourbaki expos\'e 797, Soci\'ete

1684: Math\'ematique de France (1996).

1685: 

1686: 

1687: \bibitem{lecture} V. Pasquier, A lecture on the Calogero Sutherland models,

1688: {\it The third Baltic Rim Student Seminar,} Saclay preprint,

1689: Spht-94060 (1994), [hep-th/9405104].

1690: 

1691: \bibitem{affine} V. Pasquier, Scattering Matrices and Affine Hecke

1692: Algebras, Schladming School 1995, Nucl.Phys.B (Proc.Suppl.) {\bf

1693: 45A},62-73(1996), [q-alg/9508002].

1694: 

1695: 

1696: \bibitem{Yang} C.N. Yang, Phys.Rev. {\bf 168}, 1920 (1968).

1697: 

1698: 

1699: 

1700: 

1701: 

1702: 

1703: \bibitem{Macdo1} I.G. Macdonald, {\it A new class of symmetric

1704: functions.} Actes {\bf$20^e$} Seminaire Lotharingien, p 131-171,

1705: Publications I.R.M.A. Strasbourg (1988), 372/S-20.

1706: 

1707: 

1708: \bibitem{stroganov} Y.G. Stroganov [math-ph/0204042].

1709: 

1710: \bibitem{Pasbratel} V. Pasquier, J.Phys.A {\bf 20}, 1229 (1987).

1711: 

1712: 

1713: 

1714: 

1715: 

1716: 

1717: 

1718: 

1719: 

1720: 

1721: \

1722: \end{thebibliography}

1723: \end{document}

1724: 

1725: